Despite decades of effort in academia and industry, it is not yet possible to build machines that can replicate many seemingly-basic human perceptual abilities. This work focuses on the problem of face identification that most of us effortlessly solve daily. Substantial progress has been made towards the goal of automatically identifying faces under tightly controlled conditions; however, in the domain of unconstrained face images, many challenges remain. We observe that the recent combination of widespread digital photography, inexpensive digital storage and bandwidth, and online social networks has led to the sudden creation of repositories of billions of shared photographs and opened up an important new domain for unconstrained face identification research. Drawing upon the newly-popular phenomenon of “tagging,” we construct some of the first face identification datasets that are intended to model the digital social spheres of online social network members, and we examine various qualitative and quantitative properties of these image sets. The identification datasets we present here include up to 100 individuals, making them comparable to the average size of members’ networks of “friends” on a popular online social network, and each individual is represented by up to 100 face samples that feature significant real-world variation in appearance, expression, and pose. We demonstrate that biologically-inspired visual representations can achieve state-of-the-art face identification performance on our novel frontal and multi-pose face datasets. We also show that the addition of a tree-structured classifier and training set augmentation can enhance accuracy in the multi-pose setting. Finally, we illustrate that the machine-readable “social context” in which shared photos are often embedded can be applied to further boost face identification accuracy. Taken together, our results suggest that accurate automated face identification in vast online shared photo collections is now feasible. / Engineering and Applied Sciences

Thesis (Ph.D)--Biomedical Engineering, Georgia Institute of Technology, 2008. / Committee Chair: Potter, Steve; Committee Member: Butera, Robert; Committee Member: DeMarse, Thomas; Committee Member: Jaeger, Dieter; Committee Member: Lee, Robert.

Thesis (Ph. D.)--Electrical and Computer Engineering, Georgia Institute of Technology, 2007. / Oliver Brand, Committee Member ; James H. McClellan, Committee Member ; Paul Hasler, Committee Chair ; Mark T. Smith, Committee Member ; David Anderson, Committee Member.

In this thesis, I describe a quantitative model that accounts for the circuits and computations of the feedforward path of the ventral stream of visual cortex. This model is consistent with a general theory of visual processing that extends the hierarchical model of (Hubel & Wiesel, 1959) from primary to extrastriate visual areas. It attempts to explain the first few hundred milliseconds of visual processing and “immediate recognition”. One of the key elements in the approach is the learning of a generic dictionary of shape-components from V2 to IT, which provides an invariant representation to task-specific categorization circuits in higher brain areas. This vocabulary of shape-tuned units is learned in an unsupervised manner from natural images, and constitutes a large and redundant set of image features with different complexities and invariances. This theory significantly extends an earlier approach by (Riesenhuber & Poggio, 1999) and builds upon several existing neurobiological models and conceptual proposals.First, I present evidence to show that the model can duplicate the tuning properties of neurons in various brain areas (e.g., V1, V4 and IT). In particular, the model agrees with data from V4 about the response of neurons to combinations of simple two-bar stimuli (Reynolds et al, 1999) (within the receptive field of the S2 units) and some of the C2 units in the model show a tuning for boundary conformations which is consistent with recordings from V4 (Pasupathy & Connor, 2001). Second, I show that not only can the model duplicate the tuning properties of neurons in various brain areas when probed with artificial stimuli, but it can also handle the recognition of objects in the real-world, to the extent of competing with the best computer vision systems. Third, I describe a comparison between the performance of the model and the performance of human observers in a rapid animal vs. non-animal recognition task for which recognition is fast and cortical back-projections are likely to be inactive. Results indicate that the model predicts human performance extremely well when the delay between the stimulus and the mask is about 50 ms. This suggests that cortical back-projections may not play a significant role when the time interval is in this range, and the model may therefore provide a satisfactory description of the feedforward path.Taken together, the evidences suggest that we may have the skeleton of a successful theory of visual cortex. In addition, this may be the first time that a neurobiological model, faithful to the physiology and the anatomy of visual cortex, not only competes with some of the best computer vision systems thus providing a realistic alternative to engineered artificial vision systems, but also achieves performance close to that of humans in a categorization task involving complex natural images. / PhD thesis

Pour pouvoir être autonomes dans un environnement complexe, les humains comme les systèmes artificiels doivent posséder un apprentissage souple et capable de s’adapter au changement. Dans cette thèse, nous nous intéressons à comment cette autonomie peut être obtenue par interactions entre les différents systèmes d’apprentissage de notre cerveau. Pour cela, nous modélisons dans une approche inspirée de la biologie le comportement de certaines des parties du cerveau impliquées dans les apprentissages répondant et opérant, et observons comment leurs interactions permettent un apprentissage flexible dans des tâches impliquant des changements comme l’extinction et le reversal. / In a complex environment, humans and artificials systems need a flexible learning system to adapt themselves to situations which can change. In this thesis, we study how autonomy can be the result of interactions between the different learning systems of our brain. In particular, in a biologically inspired approach, we model different parts of the brain involved in respondant and operant conditioning, et show how their interactions can promote flexible learning in tasks in which situation can change, like extinction or reversal.

How can science possibly understand the organ through which the Universe knows itself? The scientific method can be used to study how electro-chemical signals represent information in the brain. However, modelling it by simulating its structures and functions is a computation- and communication-intensive task. Whilst supercomputers offer great computational power, brain-scale models are challenging in terms of communication overheads and power consumption. Dedicated neural hardware can be used to enhance simulation performance, but it is often optimised for specific models. While performance and flexibility are desirable simulation features, there is no perfect modelling platform, and the choice is subordinate to the specific research question being investigated. In this context SpiNNaker constitutes a novel parallel architecture, with communication and memory accesses optimised for spike-based computation, permitting simulation of large spiking neural networks in real time. To exploit SpiNNaker's performance and reconfigurability fully, a neural network model must be translated from its conceptual form into data structures for a parallel system. This thesis presents a flexible approach to distributing and mapping neural models onto SpiNNaker, within the constraints introduced by its specialised architecture. The conceptual map underlying this approach characterizes the interaction between the model and the system: during the build phase the model is placed on SpiNNaker; at runtime, placement information mediates communication with devices and instrumentation for data analysis. Integration within the computational neuroscience community is achieved by interfaces to two domain-specific languages: PyNN and Nengo. The real-time, event-driven nature of the SpiNNaker platform is explored using address-event representation sensors and robots, performing visual processing using a silicon retina, and navigation on a robotic platform based on a cortical, basal ganglia and hippocampal place cells model. The approach has been successfully exploited to run models on all iterations of SpiNNaker chips and development boards to date, and demonstrated live in workshops and conferences.

京都大学 / 新制・課程博士 / 博士(工学) / 甲第23505号 / 工博第4917号 / 新制||工||1768(附属図書館) / 京都大学大学院工学研究科機械理工学専攻 / (主査)教授 松野 文俊, 教授 泉田 啓, 教授 小森 雅晴 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DGAM

This thesis deals with biologically-inspired interactive neural networks for the task of object recognition. Such networks offer an interesting alternative approach to traditional image processing techniques. Although the networks are very powerful classification tools, they are difficult to handle due to their bidirectional interactivity. It is one of the main reasons why these networks do not perform the task of generalization to novel objects well. Generalization is a very important property for any object recognition system, as it is impractical for a system to learn all instances of an object class before classifying. In this thesis, we have investigated the working of an interactive neural network by fine tuning different structural and algorithmic parameters.  The performance of the networks was evaluated by analyzing the generalization ability of the trained network to novel objects. Furthermore, the interactivity of the network was utilized to simulate focus of attention during object classification. Selective attention is an important visual mechanism for object recognition and provides an efficient way of using the limited computational resources of the human visual system. Unlike most previous work in the field of image processing, in this thesis attention is considered as an integral part of object processing. Attention focus, in this work, is computed within the same network and in parallel with object recognition. As a first step, a study into the efficacy of Hebbian learning as a feature extraction method was conducted. In a second study, the receptive field size in the network, which controls the size of the extracted features as well as the number of layers in the network, was varied and analyzed to find its effect on generalization. In a continuation study, a comparison was made between learnt (Hebbian learning) and hard coded feature detectors. In the last study, attention focus was computed using interaction between bottom-up and top-down activation flow with the aim to handle multiple objects in the visual scene. On the basis of the results and analysis of our simulations we have found that the generalization performance of the bidirectional hierarchical network improves with the addition of a small amount of Hebbian learning to an otherwise error-driven learning. We also conclude that the optimal size of the receptive fields in our network depends on the object of interest in the image. Moreover, each receptive field must contain some part of the object in the input image. We have also found that networks using hard coded feature extraction perform better than the networks that use Hebbian learning for developing feature detectors. In the last study, we have successfully demonstrated the emergence of visual attention within an interactive network that handles more than one object in the input field. Our simulations demonstrate how bidirectional interactivity directs attention focus towards the required object by using both bottom-up and top-down effects. In general, the findings of this thesis will increase understanding about the working of biologically-inspired interactive networks. Specifically, the studied effects of the structural and algorithmic parameters that are critical for the generalization property will help develop these and similar networks and lead to improved performance on object recognition tasks. The results from the attention simulations can be used to increase the ability of networks to deal with multiple objects in an efficient and effective manner.

In recent years there has been significant research in the field of computational neuroscience and many of these biologically inspired cognitive models are based on the theory of operation of mammalian visual cortex. One such model of neocortex developed by George & Hawkins, known as Hierarchical Temporal Memories (HTM), is considered for the research discussed here. We propose a simple hierarchical model that is derived from HTM. The aim of this work is to evaluate the hardware cost and performance against software based simulations. This work presents a detailed hardware implementation and analysis of the derived hierarchical model. We show that these networks are inherently parallel in their architecture, similar to the biological computing, and that parallelism can be exploited by massively parallel architectures implemented using reconfigurable devices such as the FPGA. Hardware implementation accelerates the learning process which is useful in many real world problems. We have implemented a complex network node that operates in real time using an FPGA. The current architecture is modular and allows us to estimate the hardware resources and computational units required to realize large scale networks in the future.

No description available.